Natural variation and dosage of the HEI10 meiotic E3 ligase control Arabidopsis crossover recombination

1The majority of eukaryotes reproduces via the meiotic cell division, where a diploid cell replicates DNA once and segregates chromosomes twice to produce tetrads of haploid gametes. Genetic diversity is generated between gametes due to independent chromosome segregation in addition to recombination between homologous chromosomes during meiotic prophase I. Despite the importance of crossovers for balanced chromosome segregation during meiosis and fertility, extensive genetic variation in recombination frequency is observed within and between species. Importantly, natural variation that modifies crossover frequency has the potential to widely influence genetic adaptation and the response to selection.

2Genetic polymorphisms that modify crossover frequency can be classified as cis- or trans-acting, according to whether they control recombination on the same chromosome or throughout the genome, respectively. Examples of human trans modifier loci include the RNF212 meiotic E3 ligase gene, which controls crossover levels, and the PRDM9 zinc finger SET domain gene, which specifies recombination hot spot locations. Polymorphisms are also known to exert local cis effects, where heterozygous polymorphisms can inhibit crossover repair of interhomolog strand invasion events. Structural variation (for example, insertions and deletions [indels], translocations, and inversions) are also associated with crossover suppression at larger physical scales. Extensive evidence for cis and trans modification of crossover frequency exists in plants, including Arabidopsis thaliana. Therefore, we sought to use high-throughput fluorescent reporter systems to measure recombination and identify trans-acting crossover modifier loci that vary between A. thaliana accessions.

3Meiotic recombination initiates via formation of DNA double-strand breaks (DSBs) by SPO11 transesterases. DSBs are resected to generate ssDNA, which is bound by the RecA-related recombinases RAD51 and DMC1. The resulting nucleoprotein filaments can then invade a replicated sister chromatid or a homologous DNA duplex to form a displacement loop. Immunostaining of Arabidopsis meiocytes for DMC1, RAD51, or the DSB-associated histone variant γH2A.X has revealed ∼100–200 foci distributed along the paired homologous chromosomes at the leptotene stage. Approximately 10 of these initiating meiotic DSBs mature into crossovers per Arabidopsis meiosis.

4Meiotic interhomolog strand invasion intermediates can follow alternative repair fates, including crossover or noncrossover, which differ in exchange of flanking markers. The majority of Arabidopsis crossovers is dependent on the conserved ZMM pathway (named after the budding yeast genes Zip1, Zip2, Zip3, Zip4, Msh4, Msh5, and Mer3), which includes SHOC1, HEI10, ZIP4, MSH4, MSH5, MER3, PTD, MLH1, and MLH3. Crossovers generated by the ZMM pathway show interference, where double crossovers are spaced farther apart than expected at random, which is detectable over the scale of megabases in Arabidopsis. The ZMM pathway is thought to stabilize interhomolog recombination intermediates, including double Holliday junctions, and promote crossover resolution. A minority of Arabidopsis crossovers is generated by a noninterfering pathway that includes MUS81. Noncrossover repair of strand invasion events is promoted by multiple nonredundant pathways that include FANCONI ANEMIA COMPLEMENTATION GROUP M (FANCM), MHF1, MHF2, FIDGETIN-LIKE1 (FIGL1), RECQ4A, RECQ4B, TOPOISOMERASE3α (TOP3α), and MSH2. The combined action of these anti-crossover pathways results in repair of ∼90% of initiating meiotic DSBs as noncrossovers.

5In this study, we identified an Arabidopsis natural genetic variation that acts in trans to control meiotic crossover frequency. Although A. thaliana is predominantly self-fertilizing, clear evidence for outcrossing exists. For example, Arabidopsis linkage disequilibrium decays rapidly over kilobase distances, and strong historical crossover hot spots are detectable. Genotyping of natural Arabidopsis populations has also revealed standing heterozygosity and evidence for local outcrossing between subpopulations. Therefore, recombination-modifying polymorphisms have had the opportunity to exert an effect on the genetic history of this species. Here we identify natural genetic polymorphisms in the HEI10 meiotic E3 ligase gene that associate with quantitative variation in crossover frequency between Arabidopsis accessions. We further show that HEI10 is highly dosage-sensitive and that transformation of additional HEI10 copies is sufficient to more than double crossover recombination throughout euchromatin. Together, this demonstrates that HEI10 is a limiting factor for interference-sensitive crossover formation in Arabidopsis.

6Genetic segregation of linked, hemizygous T-DNAs expressing different colors of fluorescent proteins in pollen or seed (fluorescent-tagged lines [FTLs]) can be used to measure Arabidopsis crossover frequency (Fig. 1;;;; ). We previously analyzed crossovers in an F2 population derived from crosses between the Col-420 subtelomeric FTL and Catania-1 (Ct) parents, which did not identify significant trans-acting recombination modifier loci. To further screen for natural crossover modifiers, we generated a Col-420×Landsberg erecta (Ler) F2 population, which showed higher mean recombination than Col-420×Ct (20.2 cM vs. 15.0 cM) (Fig. 1A; Supplemental Tables S1, S2) and significantly greater variation in crossover frequency between individuals (Brown-Forsythe test, P = 2.91 × 10−32). This is consistent with the presence of trans modifier loci (Fig. 1A; Supplemental Tables S1, S2).

7To identify trans recombination modifier loci, we first used Col/Ler CSLs (Fig. 1B; ). For example, CSL LCCCC denotes Ler (L) and Col (C) genotypes for each of the five chromosomes (Fig. 1B). Fourteen CSLs were crossed to Col-I2f and Col-420 FTLs (crossover reporters located on chromosomes 2 and 3, respectively), and replicate F1 measurements were collected (Fig. 1B–F; Supplemental Tables S3, S4). We observed that all F1 genotypes that were chromosome 1 Col/Ler heterozygous showed significantly reduced crossovers compared with control Col-420×CCCCC F1 plants, with weaker effects detected from the other chromosomes (Fig. 1E,F; Supplemental Tables S3–S5). This reveals the presence of a semidominant trans-acting recombination modifier on chromosome 1.

8We observed previously that juxtaposition of homozygous and heterozygous regions can influence recombination in cis at the megabase scale. To eliminate cis effects and specifically map trans recombination modifiers, we generated an F2 population from a Col-420×LLCLL cross. In this population, chromosome 3 is Col/Col homozygous, which is where the 420 FTL interval is located, and therefore cis effects were excluded. We identified two major trans rQTLs on chromosomes 1 and 4, with logarithm of the (base 10) odds ratio (LOD) scores of 40.2 and 53.5, which explain 23.3% and 33.6% of the variance in recombination, respectively (F-test, P < 2 × 10−16) (Fig. 2A; Supplemental Table S6). rQTL1Ler genotypes associate with low recombination, with heterozygotes showing intermediate crossover frequency (Fig. 2B), consistent with the semidominant effects observed for chromosome 1 in the CSL F1 experiments (Fig. 1E,F; Supplemental Tables S3, S4). In contrast, rQTL4Ler associates with high recombination and behaves recessively, explaining why it was not detected in the CSL experiments (Fig. 2C).

9To investigate the influence of rQTL1 and rQTL4 on meiotic recombination elsewhere in the genome, we performed cytogenetic analysis in Col and Ler in addition to recombinants with low (rQTL1Ler rQTL4Col) or high (rQTL1Col rQTL4Ler) 420 crossovers (Supplemental Fig. S1A–E; Supplemental Tables S7–S9). MLH1 foci occurring along the meiotic synaptonemal complex (visualized by ZYP1 immunostaining) serve as a measure of total interfering crossovers per nucleus. We observed significantly more MLH1 foci in rQTL1Col rQTL4Ler lines compared with the other genotypes (Mann-Whitney-Wilcoxon test, P = 0.0396) (Supplemental Fig. S1A,C; Supplemental Table S7). We confirmed the same trend via analysis of chiasmata at metaphase I (Mann-Whitney-Wilcoxon test, P = 2.20 × 10−5) (Supplemental Fig. S1B,D; Supplemental Table S8; ). These analyses confirm that Col and Ler polymorphisms underlying rQTL1 and rQTL4 influence crossovers not only in the 420 interval but throughout the chromosomes.

10We sought to identify rQTL1 using an F2 population derived from a Col-420×LCCCC cross, which again revealed a major rQTL on chromosome 1 (Fig. 2D; Supplemental Table S10). We selected an F2 individual that was Col/Ler heterozygous spanning rQTL1 (19.54–21.24 Mb), which was self-fertilized to generate a large F3 population (n = 3072) (Fig. 2E; Supplemental Table S11). Genotyping identified 307 F3 plants with crossovers within the rQTL1 region, which were then measured for 420 crossover frequency and genotyped for 15 additional markers (Fig. 2E; Supplemental Table S11). This narrowed the credible rQTL1 interval to a 34-kb region containing 14 genes (Fig. 2E,F). The most strongly associated marker pair (LOD = 51.02) defined a 4.3-kb interval containing two genes: MRD1 (At1g53480) and HEI10 (At1g53490) (Fig. 2F). As HEI10 belongs to a conserved gene family, which encodes RING domain SUMO/ubiquitin E3 ligases that promote crossovers in diverse eukaryotes, this was the strongest candidate gene for rQTL1 (Supplemental Figs. S2, S3;;;;;;;;;; ). HEI10 family proteins possess N-terminal RING domains, central coiled-coil domains, and C-terminal regions of unknown function (Supplemental Figs. S2, S3B; ).

11To further investigate HEI10 polymorphisms associated with rQTL1, we sequenced the Ler accession used in our experiments and identified a single nonsynonymous (R264G) substitution and three synonymous intragenic variants relative to the Col reference sequence (Fig. 3A; Supplemental Table S12), which was consistent with 1001 Genomes project data from the closely related Ler-1 and La-0 accessions. The HEI10 promoter, which overlaps the antisense gene MRD1, is also polymorphic, with 26 single-nucleotide polymorphisms (SNPs) or indels upstream of the start codon (Fig. 3A; Supplemental Table S12). We generated additional F2 populations derived from Col-420×Bur-0 or Col-420×Cvi-0 crosses and again observed significant association between the HEI10 region and crossover frequency (Bur-0 LOD = 8.30, 95% significance threshold LOD = 2.82; Cvi-0 LOD = 31.74, 95% significance threshold LOD = 2.97) (Fig. 3B,C; Supplemental Table S13). We sequenced HEI10Bur and HEI10Cvi and observed 43 and 30 polymorphisms, respectively, relative to HEI10Col (Fig. 3A; Supplemental Table S12). As we previously observed an absence of trans rQTLs in Col-420×Ct-1 populations (Fig. 3D; ), we also sequenced HEI10Ct (Fig. 3A; Supplemental Table S12). Twelve polymorphisms are shared between HEI10Ler, HEI10Cvi, and HEI10Bur but absent from HEI10Ct, including the R264 substitution, which we consider as candidates for rQTL1 causal variants (Fig. 3A; Supplemental Table S12). HEI10 transcript and protein levels measured by quantitative RT–PCR (qRT–PCR) and immunocytogenetic analysis, respectively, did not show significant differences between Col, Ler, and Col/Ler F1 (Supplemental Fig. S1E; Supplemental Tables S14–S16), consistent with the causal rQTL1 polymorphism influencing HEI10 protein function or expression timing rather than expression level. The nonsynonymous R264G polymorphism occurs in the HEI10 C-terminal region (Fig. 3E; Supplemental Figs. S2, S3), which, by analogy with other RING E3 ligases, may play a role in substrate recognition. We queried 1001 Genomes project data for the frequency and geographic distribution of the HEI10 R264G variants (Supplemental Fig. S4; ). Both alleles are globally distributed, with the majority of accessions (959 out of 1008) showing the Ler G264 genotype, and the Col R264 reference allele present in the remaining 123 (11.4%) (Supplemental Fig. S4). As other Brassicaceae HEI10 orthologs show glycine at position 264, this is consistent with the R264 variant observed in Col-0 and Ct-1 being more recently derived within A. thaliana (Fig. 3E).

12As a genetic test of rQTL1 allelism with HEI10, we crossed recombinant 420 lines with the null hei10-2 allele, which was isolated in the Col background and shows substantially reduced crossovers and fertility when homozygous. Four independent recombinant lines were used for crosses, two of which were rQTL1Col homozygous, and two that were rQTL1Ler homozygous (Supplemental Table S17). All lines analyzed were rQTL4Ler homozygous. F1 individuals that were hei10-2 heterozygous showed significantly reduced recombination compared with wild-type crosses (Fig. 4A; Supplemental Table S17). This indicates HEI10 dosage sensitivity, which is similar to haploinsufficiency of mouse hei10 and rnf212 mutations. Consistent with our previous rQTL mapping, the progeny from rQTL1Ler crosses showed significantly fewer 420 crossovers compared with rQTL1Col progeny (X2, P < 2.2 × 10−16) (Fig. 4A; Supplemental Table S17), which we interpret as reflecting the different activity of HEI10 Col/Ler variants, including R264G. To investigate whether haploinsufficiency is a general property of mutants in the Arabidopsis ZMM pathway, we compared Col-420 F1 crossover measurements using hei10-2, msh4-1, msh5-1, shoc1-1, and ptd1 heterozygotes (Fig. 4B; Supplemental Table S18). Among these mutations, only hei10-2/+ heterozygotes showed significantly reduced crossovers compared with wild type (X2, P = 1.63 × 10−38) (Fig. 4B; Supplemental Table S18), revealing that dosage sensitivity was specific to HEI10.

13Due to HEI10 dosage sensitivity, we next investigated whether increasing copy number would elevate recombination beyond wild type (Fig. 4C–E). We transformed Col-420 plants with a HEI10 transgene under the control of its endogenous promoter, amplified from either Col or Ler genomic DNA (referred to here as HEI10Col and HEI10Ler) (Fig. 4C–E). HEI10Col and HEI10Ler T1 populations, but not empty vector T1, showed significantly higher recombination than untransformed Col-420 controls (Mann-Whitney-Wilcoxon test: HEI10Col, P = 4.03 × 10−14; HEI10Ler, P = 3.64 × 10−10; empty, P = 0.474) (Fig. 4E; Supplemental Tables S2, S19). Wide variation in recombination rate was observed within T1 populations (Fig. 4E), which was likely caused by varying HEI10 transgene copy numbers and position effects that influence expression level. Indeed, qRT–PCR analysis of HEI10 expression from meiotic stage flower buds (floral stages 1–12) of T1 transformants revealed a positive correlation with 420 recombination (r = 0.727; P = 7.96 × 10−6) (Fig. 4F; Supplemental Table S20). HEI10Col transformants showed higher recombination than HEI10Ler transformants (Mann-Whitney-Wilcoxon test, P = 8.93 × 10−3) (Fig. 4E; Supplemental Table S19), which is consistent with the different recombination activity of Col/Ler HEI10 variants. To further investigate polymorphisms responsible for differences in HEI10 function, we generated Col::Ler and Ler::Col promoter swap constructs and repeated transformation (Fig. 4E; Supplemental Table S19). The HEI10-Ler::Col transformants showed significantly higher recombination than HEI10-Col::Ler transformants (Mann-Whitney-Wilcoxon test, P = 1.1 × 10−2). Furthermore, HEI10-Ler::Col were not significantly different from HEI10-Col::Col (Mann-Whitney-Wilcoxon test, P = 0.841), and HEI10-Col::Ler were not different from HEI10-Ler-Ler (Mann-Whitney-Wilcoxon test, P = 0.259). Together, this is consistent with intragenic HEI10 polymorphisms, including R264G, causing differences in recombination activity (Fig. 4E; Supplemental Table S19).

14A HEI10Col T1 line showing high 420 recombination (C2; 33.74 cM) was selected for cytological investigation. Immunostaining of leptotene stage meiotic nuclei for HEI10 showed a significant increase in signal intensity (Mann-Whitney-Wilcoxon test, P = 3.90 × 10−4), although focus numbers were not changed (Mann-Whitney-Wilcoxon test, P = 0.5971) (Fig. 5A–C; Supplemental Tables S21, S22). To investigate the effect of HEI10Col transformation on crossover formation, we performed MLH1 immunostaining at the pachytene stage (Fig. 5D,E). There were close to double the number of MLH1 foci along HEI10Col chromosomes compared with wild type (mean = wild type 9.3, HEI10Col 16.2; Mann-Whitney-Wilcoxon test, P = 4.83 × 10−8) (Fig. 5D,E; Supplemental Table S23). HEI10Col also showed more compact bivalents at metaphase I, which is indicative of greater crossover numbers in the chromosome arms (Fig. 5D; ). This provides cytological evidence that increased HEI10 dosage and expression level elevates crossovers throughout the genome.

15To investigate the effect of increased HEI10 dosage on crossovers at higher resolution, we used genotyping by sequencing (Fig. 6A). The HEI10Col C2 line was backcrossed to Ler alongside a wild-type Col control (Fig. 6A). HEI10Col×Ler F1 plants showed highly elevated 420 recombination compared with Col/Ler F1 (Supplemental Tables S19, S24), demonstrating that HEI10 increases crossovers in both hybrid (Col/Ler) and inbred (Col/Col) backgrounds. F2 populations (n = 192) were then generated from wild-type and HEI10Col Col/Ler F1 plants and sequenced to identify crossover locations (Fig. 6A; Supplemental Table S25; ). The HEI10Col population contained more than double the number of wild-type crossovers (1230 vs. 2928 crossovers; mean per F2, 6.41 vs. 15.25; Mann-Whitney-Wilcoxon test, P = 1.07 × 10−59) (Fig. 6B; Supplemental Fig. S5; Supplemental Table S1), consistent with our MLH1 focus analysis (Fig. 5D,E). Euchromatic chromosome arms showed the greatest increase in HEI10Col crossovers (2.6×, Mann-Whitney-Wilcoxon test, P = 1.98 × 10−8), with the largest effects in the subtelomeric regions (Fig. 6C,D; Supplemental Tables S26, S27). A lower yet significant crossover increase was observed in the pericentromeres (1.6×, Mann-Whitney-Wilcoxon test, P = 2.34 × 10−9) (Fig. 6C; Supplemental Fig. S6; Supplemental Tables S26, S27). In contrast, centromeric suppression of crossovers was observed in both populations (Fig. 6C; Supplemental Fig. S6; Supplemental Tables S26, S27;; ).

16Crossovers were mapped using ∼1×–2× depth sequencing data and the TIGER analysis pipeline (Supplemental Table S25; ), which resolved events to a mean width of 976 base pairs (bp). To analyze the fine-scale distribution of wild type versus HEI10 crossovers, we overlapped them with gene and transposon annotations and compared them with matched sets of randomly chosen intervals (Supplemental Fig. S6A; Supplemental Table S28). Both wild-type and HEI10 crossovers show increased intergenic and decreased transposon overlap compared with random (Supplemental Fig. S6A; Supplemental Table S28), which is consistent with Arabidopsis crossover hot spots associating with euchromatic gene promoters and terminators. We also compared DNA methylation levels and observed that crossovers from both populations were hypomethylated in CG, CHG, and CHH sequence contexts compared with random (Supplemental Fig. S6B; ). This is further consistent with both wild-type and HEI10 crossovers being enriched within euchromatic regions along the chromosome arms.

17Plants, fungi, and invertebrates possess a single HEI10/RNF212 ortholog, whereas vertebrates encode separate RNF212 and HEI10 proteins that function as antagonistic SUMO and ubiquitin E3 ligases. In mice, HEI10 and RNF212 promote stable accumulation of MSH4/MSH5 (MutSγ) heterodimers on recombining meiotic chromosomes via a SUMO–ubiquitin relay, which promotes recruitment of MLH1/MLH3 (MutLγ) and crossover formation. However, it remains unclear which proteins are direct ubiquitin/SUMO targets of the HEI10/RNF212 E3 ligases, although rice HEI10 can directly interact with MSH5. HEI10/RNF212 RING domains are C3HC4 zinc fingers, which generally function as protein interaction domains to recruit E2 ubiquitin-conjugating enzymes to substrates. Regions outside the RING domain are known to contribute to substrate recognition. Therefore, we postulate that the HEI10 C-terminal R264 variant may alter substrate recognition efficiency and SUMO/ubiquitin transfer during regulation of meiotic recombination.

18Beyond genetic variation that alters HEI10 function, we demonstrated that Arabidopsis crossover frequency is exquisitely sensitive to HEI10 dosage. We propose that higher HEI10 concentration at meiotic repair foci quantitatively promotes crossovers via increased SUMO or ubiquitin transfer to substrate recombination factors. The dosage sensitivity of Arabidopsis HEI10 is strikingly reminiscent of rnf212 and hei10 mutations in mice, which show haploinsufficiency. Furthermore, polymorphisms in RNF212 and HEI10 genes have been associated with variation in recombination rate in human, cattle, and sheep populations. We propose that haploinsufficiency and dosage sensitivity of HEI10/RNF212 genes predisposes them to acting as trans recombination modifiers in diverse eukaryotic lineages. It is interesting to note that increased HEI10 dosage in Arabidopsis led to the greatest crossover increase in subtelomeric euchromatin, which is similar to the sex differences in recombination observed in both plants and mammals. For example, Arabidopsis male meiosis shows subtelomeric increases in crossover frequency. Therefore, we speculate that differences in HEI10/RNF212 expression or regulation have the potential to contribute to sex differences in recombination. We also note that increasing HEI10 copy number may be an attractive mechanism to elevate crossover numbers during breeding of crop species.

19Crossover modifier loci are able to alter population responses to selection. For example, recombination can mitigate the effects of Hill-Robertson interference when linked loci are under selection. Therefore, loci that modify crossover frequency may influence genetic adaptation to diverse environments and conditions. Interestingly, total recombination levels compared across eukaryotes are generally low, typically with one or two crossovers per chromosome per meiosis, despite wide variation in physical genome size. It is possible that high recombination levels might cause infertility and be selected against. However, Arabidopsis anti-crossover pathway mutants show normal fertility despite greatly elevated crossover frequency, at least in the short term. Therefore, we propose that Arabidopsis recombination modifiers may act to maintain relatively low crossover numbers. As rQTL1Col and rQTL4Col alleles show opposite effects on crossover frequency, this example is consistent with antagonistic modifiers acting to balance recombination. It is also important to note that the effect of modifiers will be highly dependent on genome architecture and outcrossing levels. Crossover modifiers may be especially common in plants, where frequent polyploidization causes challenges for balanced meiotic genome transmission. Indeed, meiotic axis proteins (ASY1, ASY3, PDS5, ZYP1a, ZYP1b, SMC1, and REC8) have been strongly selected during polyploid evolution in Arabidopsis arenosa, and the Ph1 locus is required for promotion of homologous versus homeologous recombination in hexaploid bread wheat. Therefore, further study of plant meiotic modifier loci is likely to reveal insights into the control of recombination and how this interacts with selection during evolution.

20Crossover frequency was measured using fluorescent reporters in seeds (Col-420) and pollen (Col-I2f). In F2 populations derived from FTL hemizygotes, only a subset of progeny will contain the fluorescent protein-encoding transgenes also in a hemizygous state, which is necessary for crossover measurement. When using the seed-based 420 line, it is possible to enrich for FTL hemizygous F2 plants by examining seed under a fluorescence microscope prior to sowing and separating nonfluorescent, hemizygous, and homozygous seeds based on eGFP and dsRed fluorescence intensities. CSLs were kindly provided by Erik Wijnker, Jose van der Belt, and Joost Keurentjes (University of Wageningen), with the exception of LCCCC, which was obtained from an esd7-1 backcross line. Mt-0, Ct-1, and Cvi-0 accessions were obtained from the Nottingham Arabidopsis Stock Centre. The ZMM mutant alleles used were hei10-2 (Salk_014624), msh4-1 (Salk_136296), msh5-1 (Salk_110240), ptd1 (Salk_ 127447), and shoc1-1 (Salk_057589). Genotyping primer sequences for these mutations used with the LBb1.3 T-DNA left border primer are in Supplemental Table S29.

21Genomic DNA was extracted using CTAB and genotyped using PCR amplification of Col/Ler SSLP, CAPS, or dCAPS markers (Supplemental Tables S30, S31). We performed one- and two-dimensional QTL mapping using the R statistical package rQTL. We implemented the Haley-Knott regression algorithm using 2.5-cM steps across the genome and 0.1-cM steps for rQTL1 fine mapping. To fit models with multiple QTLs, we used the fitqtl function with Haley-Knott regression. We used 10,000 permutations for each mapping population to empirically calculate genome-wide LOD score significance thresholds.

22HEI10 was amplified from Col or Ler genomic DNA using primers HEI10-XbaI and HEI10-BamHI (Supplemental Table S29). Amplification products were cloned into the pGREEN-0029 binary vector using XbaI and BamHI restriction enzymes. Promoter swap constructs were generated using XbaI/PacI digestion and vector religation. These vectors were transformed into Col-420 FTL hemizygous plants using Agrobacterium strain GV3101 and floral dipping.

23RNA was extracted from ∼40 mg of immature flower buds (closed buds up to stage 12, which contain all meiotic stages) using TRI reagent (Sigma-Aldrich). Reverse transcription was performed with SuperScript II reverse transcriptase (ThermoFisher Scientific). Relative HEI10 expression was measured by qPCR using primers HEI10-qPCR1 and HEI10-qPCR2, and the meiosis-specific gene DMC1 was amplified using primers DMC1-qPCR1 and DMC1-qPCR2 as a control for ΔCt calculations (Supplemental Table S29). For HEI10 T1 analysis, the 2−ΔΔCt method was used to quantify relative transcript levels in comparison with untransformed plants.

24Chromosome spreads of Arabidopsis pollen mother cells and rDNA in situ hybridization were performed as described using fixed buds. Pachytene stage meiocytes were immunostained for ASY1, ZYP1, MLH1, and HEI10 using fresh buds as described with the following antibodies: α-ASY1 (rabbit; 1:500 dilution), α-MLH1 (rabbit, IgG-purified; 1:200 dilution), α-ZYP1 (rat; 1:500 dilution), and α-HEI10 (rabbit; 1:200 dilution). HEI10-immunostained slides within experiments were prepared side by side, and images were captured using the same exposure times. HEI10-immunostained cell images were acquired as Z-stacks of 10-µm × 0.4-µm optical sections, and maximum intensity projections were reconstructed using ImageJ. Cell boundaries were defined manually, and total signal intensity within cells was measured. An adjacent image region was used to measure background intensity, and this value was subtracted from the cell intensity. The fluorescence signal from an adjacent Inspeck Red microsphere (ThermoFisher Scientific) was also used to normalize HEI10 signal intensity. Microscopy was conducted using a DeltaVision personal DV microscope (Applied Precision/GE Healthcare) equipped with a CDD Coolsnap HQ2 camera (Photometrics). Image capture was performed using SoftWoRx software version 5.5 (Applied precision/GE Healthcare). For MLH1 and ZYP1 coimmunostaining of pachytene nuclei, individual cell images were acquired as Z-stacks of 0.1-µm optical sections, and the maximum intensity projection for each cell was rendered using ImageJ. Numbers of MLH1 foci associated with the synaptonemal complex were scored. DAPI staining of chromosomes from metaphase I nuclei and chiasma counting were performed as described. Image capture was conducted using a Nikon 90i fluorescence microscope. Images were analyzed with NIS-Elements-F software and ImageJ.

25DNA was extracted using CTAB and used to generate sequencing libraries as described with the following modifications. DNA was extracted from three rosette leaves of 5-wk-old plants, and 150 ng of DNA was used as input for each library. DNA was sheared for 20 min at 37°C with 0.4 U of DNA Shearase (Zymo research). Each set of 96 libraries was sequenced on one lane of an Illumina NextSeq500 instrument (300-cycle Mid Output run). FastQ sequencing data files are available from ArrayExpress accessions E-MTAB-4657 (wild type) and E-MTAB-4967 (HEI10). Sequencing data were analyzed to identify crossovers as reported previously using the TIGER pipeline. Crossovers were tallied in 10-kb windows and plotted along chromosomes after smoothing using the R function “filter.” Crossovers were counted and compared between populations using 2 × 2 contingency tables and χ2 tests.